Beyond diagonal-IRS assisted UAVs in terahertz network utilizing twin-delayed deep deterministic policy gradient approach

This paper presents an innovative approach in Terahertz (THz) band communications in 6G network, utilizing the integration of a Beyond Diagonal Intelligent Reflective Surface (BD-IRS) and Unmanned Aerial Vehicle (UAV) communications. This work examines the significance of THz downlink communication system in achieving an optimal trade-off between system performance and the circuit topology complexity of BD-IRS. An optimization framework named a (BD-IRS UAV) that optimizes the beyond diagonal BD-IRS phase shift in the reflective mode and finds the optimal UAVs locations, aiming to maximize the system data rate has been proposed. The multi BD-IRS UAV formulated problem is modeled as a Markov Decision Process. Nevertheless, this optimization problem is a non-convex and it is difficult to solve optimally. To tackle this difficulty, a Deep Reinforcement Learning based framework utilizing Twin-Delayed Deep Deterministic Policy Gradient (TD3) algorithm is proposed to achieve an efficient solution. Finally, simulation results demonstrate the emphasis for the proposed robust algorithm-assisted THz system which significantly improves the system data rate and impacts on performance of the system when compared with the benchmark approaches.